Anatomical data
Visual Reports MRIQC unprocessed T1w data
In this document we will make notes on how to interpret the visual reports of MRIQC specifically for T1w data.
Common artifacts with bold data
For:
Susceptibility distortion artifacts
- Signal drop-out
- Brain distortions
Description is the same as the BOLD reports.
Note that normalization and co-registration are relative robust to structural images with mild artifacts, therefore it is not always absolute necessity to impose exclusion criteria on the unprocessed T1w images.
Examples:
Intensity non-uniformity:
Is characterized by a smooth variation (low spatial frequency) of intensity throughout the brain caused by the strongel signal sensed in the proximity of coils.
- Where: On the zoomed-in view on the T1w image
- Can be a problem for automated processing methods that assume a type of tissue [e.g. white matter (WM)] is represented by voxels of similar intensities across the whole brain.
- An extreme intensity non-uniformity can also be a sign of coil failure
For an example see the above image Figure S10F
Eye spillover
Eye movements may trigger the signal leakage from the eyes through the imaging axes with the lowest bandwidth (i.e., acquired faster), potentially overlapping signal from brain tissue.
- A strong signal leakage can however be noticeable on the zoomed-in view of the T1w image
- In defaced data the leakage might not be visible
For an example see the above image Figure S10G
Other Notes on Interpretation
Interpretation of the visual reports of the T1w images based on (“Introduction to MRIQC [TRAIN-05-2022] - YouTube,” n.d.)
Zoomed-in mosaic view of the brain
Zoomed in brain masks of the T1w images in a horzontal view
We want high contrast between the grey matter and the white matter
We also have to check for ringing artifacts
Background noise
Yellow and green are areas with high background noise
Dark purple are areas with lwo background noise
Areas like:
The teeth
Sinus cavities
Air canals
- Are areas with high background noise
Example:
But: whenever you get in the brain you don’t really want any noise, you want that all purple
Example:
And same thing for the saggital view
Example:
(Note: After discussing with Gaëlle, the above examples are super super clean, and is not that usual to actually get so clean data, it’s ok).